Julie Zhuo announced the sale the way founders usually do, with a short post and a photograph. “Nearly six years ago, Chandra and I started Sundial,” she wrote on X. On Friday, she said, the company had been acquired by Meta, and she and her co-founder Chandra Narayanan would be joining along with much of the team.
Axios reported the deal the same day, and a Meta spokesperson confirmed it. The company did not disclose the price. Sundial built an AI-powered data analytics platform, the kind that lets an employee type a question in plain English and get back a number, a chart or a sentence explaining why a metric moved. It sold to businesses that wanted their analysts to stop being a ticket queue.
The unusual part is where the founders are going. Zhuo spent more than a decade at Facebook and rose to vice president of product design before leaving in 2020; her book on management found a wide audience in Silicon Valley. Narayanan was director of analytics at Facebook, then head of analytics at Instagram and at Sequoia Capital, where he spent years looking at the metrics of startups that wanted Sequoia’s money. Both are returning to a company they know better than most of the people who stayed.
They will land inside DATA, the unit Meta created in July when it named Alex Schultz its first chief data officer. Schultz, previously the company’s marketing chief, described the organization on LinkedIn as responsible for descriptive analysis, for finding questions the company has not yet thought to ask, for running experiments on its data and AI products, and for monitoring the company’s critical AI systems and raising alerts when something drifts.
His post made the rationale explicit. “AI has created effectively unlimited demand for insights, and we want to define the future of AI analytics,” Schultz wrote. He said Narayanan would lead Meta’s work on analytics tooling and agents.
That phrase, tooling and agents, is the part worth paying attention to. Meta employs tens of thousands of people who generate questions about its products, and the answers historically came from data scientists who spent their weeks writing queries. An agent that can translate a question into a correct query, check its own work and explain the result changes the ratio of questions to analysts. It also changes the failure mode: a wrong number delivered confidently is worse than a slow one delivered late, which is why the vendors selling this category spend most of their engineering effort on semantic layers and permissioning rather than on the language model.
Sundial’s pitch was aimed at exactly that gap. Its chatbot plugin let teams across an organization query company data without routing every request through a central analytics group, which is the bottleneck every large company recognizes. Whether that architecture survives contact with Meta’s internal systems is an open question, and the answer will not be public. Sundial will stop serving outside customers when the transaction closes, so the product as it existed last week disappears.
The deal fits a pattern that has defined the year. Large technology companies have been buying small AI companies for teams rather than for revenue, often at prices that make sense only if the acquirer expects the people to build something it cannot hire for. Meta has been among the most aggressive, paying extraordinary sums for researchers and restructuring its AI organization twice in eighteen months. An analytics startup with a clever interface is a rounding error against that spending, but it addresses a different problem: the company’s own ability to understand what its models are doing.
There is a competitive logic as well. The analytics layer is where enterprise AI budgets are actually being spent, because it is the one use case with a measurable return. Databricks and Snowflake have built their roadmaps around it, Microsoft has pushed it into Fabric, and a long list of startups has raised money on the promise of replacing the analyst. Meta is not selling that software to anyone. It is buying the capability for itself, which tells you something about how confident large companies are that the tooling will be commoditized.
For Zhuo, the return closes a loop that is common in Silicon Valley and rarely this neat. She wrote publicly for years about design, management and the difficulty of measuring work that produces no artifact. Sundial was an attempt to make one kind of measurement easier. Selling it back to the company where she learned the trade, to a unit run by a colleague from the marketing org, is the sort of outcome founders describe as a fit rather than an exit.
Meta has not said what it paid, how many employees are joining or what will happen to Sundial’s customers beyond the shutdown. It does not need to. The hire is the announcement, and the message to Meta’s data organization is that the company intends to build its analytics agents in-house rather than rent them.
